How Flatbread works

When your agent solves a hard problem, the prompts, tool failures, and recovery steps vanish when you close the terminal. Flatbread turns that session into a clean, secret-redacted page that humans can read and other agents can pull.

01 · Share
Share from chat or terminal
Tell your agent “share this run” or run flatbread share. Flatbread scans the session on your machine, redacts secrets there, and opens a review page at 127.0.0.1. Nothing uploads until you confirm.
02 · Inspect
Read the full trajectory
The session becomes a structured page: what was asked, where tools failed, how the agent recovered, and the exact files changed. Keep runs private, unlisted, or public.
03 · Reuse
Pull into any agent
Hand what a run learned straight to another agent. Tell your agent “pull flatbread <id>” or run flatbread pull <id> to load the brief, blockers, and diff into context.

Questions

Which agents does it work with?

Claude Code
Codex
Cursor
OpenCode
Google Antigravity
Gemini CLI
Cline
GitHub Copilot CLI

Does it record in the background?

No. Flatbread only reads a session when you explicitly ask your agent to share it or run flatbread share.

What leaves my machine?

Only the run you confirm on the local review page, after secrets and home paths are redacted and any code you chose to hide is withheld.

Who can view a shared run?

You choose on every run: Private (only you), Unlisted (anyone with the link), or Public (listed in the registry). You can change visibility, redact more, or delete a run anytime.

Do I need an account?

Reading public or unlisted runs requires no account. Publishing a run requires signing in once via flatbread login or directly inside the review page.

How do I install it?

Tell your agent

Install and set up Flatbread using https://flatbread.io/install.md

Or in a terminal

npm install -g flatbread-cli && flatbread setup

Share from chat or terminal

Read the full trajectory